Till date, eight of the 84 winners of the Prize in Economic Sciences in Memory of Alfred Nobel have received it for work on development economics: Simon Kuznets (1971), Gunnar Myrdal (1974), Theodore Schultz and W Arthur Lewis (1979), Angus Deaton (2015), and Abhijit V Banerjee, Esther Duflo and Michael Kremer (2019). Banerjee’s, Duflo’s and Kremer’s studies are based on detailed sampling work through randomised control trials (RCTs).
What are RCTs? Starting in 1948, with testing the use of streptomycin for treating pulmonary tuberculosis, RCTs have been standard tools to evaluate whether a drug works or not. An RCT randomly allocates volunteers to, say, two groups. One is treated with the drug in question, the other with a placebo. There are over 200,000 well-documented cases of RCTs in medicine.
Can RCTs be designed to ask micro questions in social science? Yes, they can, and have been by Banerjee, Duflo, Kremer and others.
Here are two examples of RCT-based research. The first is on primary education in Kenya (Duflo, Pascaline Dupas and Kremer, Journal of Public Economics, 2015). There are two types of schoolteachers in Kenya: government employed, and those hired locally by parent teacher associations (PTAs). The sarkari teachers are protected and get much higher emoluments compared to the others.
In this experiment, the PTAs were funded to hire qualified graduates of teacher training colleges as extra contract teachers for two years. They were paid a quarter of what government teachers got, and their contracts could be renewed after a year, but only after PTA approval. Contract teachers followed the same curriculum as government teachers, and were assigned to teach a fixed group of first-grade students.
What did the study find? First, providing school PTAs funds to hire extra contract teachers improved learning. Test scores of students in such schools were 0.22 standard deviations higher than those of students in the control schools. Moreover, students who were exclusively assigned to contract teachers enjoyed a 0.29 standard deviation gain in scores, while those under sarkari teachers did not.
As Easy as ABC
Second, in schools with only government teachers, the ‘bunking’ rate was 42%. In contrast, despite their lower pay, contract teachers were 20% more likely to be teaching during random visits.
As a result, their students learnt more. Moreover, sarkari teachers in schools with contract teachers decreased their own efforts: the probability that they would be found in class teaching during a random visit decreased by 27%. Not surprisingly, their students did not significantly gain in learning.
What does this RCT suggest? That instead of dealing with the teaching problem by appointing more government teachers, it makes sense to empower local parents’ bodies to appoint capable contract teachers. It’s less expensive and more effective, since contract teachers have better incentives to teach. Can it be replicated elsewhere? Yes, if there is an adequate supply of qualified contract teachers, and if sarkari teachers cannot politically prevent outsiders.
The second example is immunisation, which remains abysmally low in ‘tribal India’. With high absenteeism among sarkari health staff, poor families are averse to taking their children to health centres to complete a fivephase immunisation schedule.
An RCT in a tribal area of Udaipur in Rajasthan had (i) a control group of villages; (ii) a second, call these ‘A villages’, which were served by a mobile immunisation team arriving at fixed days and times in different villages; and (iii) athird, call these ‘B villages’, where, in addition to mobile teams, each immunisation was rewarded with a kilogram of lentils and, after all five, a set of thalis.
In the control group, only 6.2% of kids in the 1-3 years age group were fully immunised; in intervention A villages, it was 16.6%; and in the B villages, it was 38.3%. Also, children from villages that neighboured B were 20% more likely to be fully immunised than others —suggesting that the news of efficient immunisation plus incentives travelled from one village to another.
And since immunisations were more regular in the B cluster, the cost per fully vaccinated child was lower than elsewhere (Banerjee, Duflo, Rachel Glennerster and Dhruva Kothari, British Medical Journal, 2010). Is this development economics? Yes.
The goals of development are to create and sustain interventions that reduce poverty, raise educational levels, ensure better health and eliminate avoidable mass illnesses and hunger. There are many ways of doing these. But before actually engaging in such intermediations, it helps to ascertain if these are properly designed. Micro-level RCTs by Banerjee, Duflo, Kremer and others give clues about design and workability.
Is development economics only about RCT-based studies? Not at all. There is scope in doing ‘big’ development economics — by asking economy-wide questions and using large data sets to present plausible answers that could fashion better interventions. If done carefully and well.
No Random Tool
Are RCTs silver bullets for all of poverty’s and development’s questions? Banerjee, Duflo or Kremer haven’t claimed that. It is a tool, and not the only one. Yes, Banerjee and Duflo speak persuasively of the tool and, in doing so, may have evangelised it more than a bit.
However, they can’t be blamed for the Royal Swedish Academy of Sciences stating that their work “has considerably improved our ability to fight global poverty” and that “their new experiment-based approach has transformed development economics”. This has got the goat of some detractors, for how can such seemingly obvious minutiae deserve the big prize and overflowing praise? Well, pity them.
(The writer is chairman, Corporate and Economic Research Group (CERG) Advisory)